Adopting Cognitive Technologies

By Agustinus Nalwan, Head of AI And Machine Learning, Carsales.Com Ltd

Agustinus Nalwan, Head of AI And Machine Learning, Carsales.Com Ltd

1. In your opinion, how have various Cognitive technology disciplines evolved over the years? What are the myriad advantages of adopting cognitive technology disciplines for businesses today?

Cognitive technology has been evolving rapidly over the years, primarily due to the vast amount of data people share (photos, blogs, and videos, etc). This trend, coupled with the computing power we now have, enables us to build more complex and powerful artificial intelligence (AI), such as speech, recommendation, and image recognition. Cognitive technology allows businesses to be more agile. For example, an insurance company could use a chat-bot and a recommendation system powered by AI to personalize an insurance quote 24/7, which may help to reduce operational costs.

2. What according to you are some of the challenges plaguing Cognitive technologies today and how can they be effectively mitigated?

I feel the biggest challenge plaguing cognitive technologies is acquiring enough data to train your AI model, due to the limitation of current AI technologies.

The AI presently being built is very good at performing a task in one knowledge domain; it’s deep but very narrow. For example, current AI fails to identify cars that it has not seen before, even if presented with a Toyota Corolla and the word “Corolla” is clearly visible on the car, which should be sufficient for the AI to guess what the car is with a high degree of accuracy. This makes it hard to train AI without lots of labeled training data, which is a problem currently experienced by many.

This can be mitigated, to a certain extent, by employing a technique called a transfer learning, which basically re-purposes already trained AI specific to your business current problem. For example, you can take an AI model that knows how to recognize fruit and animals, then train it further to recognize a car. This will considerably help lower the training set requirements.

Another way to mitigate the data scarcity is to transform your training data into a more meaningful representation before you feed it into your AI model. This will help the AI by supplying information from other knowledge disciplines.

Businesses won’t just adopt AI technology to stay ahead of their competitors; they will adopt it to stay relevant

Despite the challenges, there are many benefits of using AI, and I believe boosting productivity has to be one of the most significant.

3. What according to you are some of the technological trends influencing Cognitive technologies today?

I think smartphones, robotics, and IoT devices are definitely the biggest influence on cognitive technologies. Almost everyone has a smartphone today, and there is a vast amount of information people can access and consume through smartphones.

The issue is to deliver the right amount of information at the right time to users, which is where AI technology be very helpful i.e. recommendations and personalization of content.

In recent years, myriads of IoT devices have been entering our lives, such as fitness trackers and smart home sensors. These devices collect lots of data, for example – calories burned and room temperature. AI technologies are required to understand the sensor data, personal preferences, and patterns so that they can take appropriate action.

4. What are some of the best cognitive technology practices enterprises should adopt today to steer ahead of competitors?

Building AI technology requires a lot of data, and generally, enterprises have a strong advantage because they have data and collect data regularly. Therefore, data capitalization should be a priority for enterprises; however, the data needs to be useable.

Good data governance is essential to ensure the data is clean so that is can be used to help build AI technologies now and in the future. For example, to identify a price trends accurately, instead of just collecting information on what price bracket the cost of a product falls in i.e. ($20k-$25k), the actual price of the product should be collected ($24,300), which then makes it possible to build AI technology that requires price data at a granular level.

5. Do you have any suggestions or a piece of advice for industry veterans or young entrepreneurs regarding the adoption of cognitive technology disciplines?

In the near future, businesses won’t just adopt AI technology to stay ahead of their competitors; they will adopt it to stay relevant. If businesses, especially enterprises, have not already implemented cognitive technology, I recommend they do so.

Most often than not, the building phase of cognitive technology won’t be the biggest hurdle businesses face. The integration process is often the most challenging, especially for a large enterprise with an already established customer base.

Trust can take time to gain within an organization, especially when adopting AI technology because it can be hard to explain. For those who are working in a corporate environment that are keen to explore implementing AI technology, I highly recommend identifying smaller projects that can make a big impact, especially because building AI solutions take time. This approach is a great way to gain trust and understanding from the business to invest more.

You may not get AI technology right on your first deployment, so start small and continuously make improvements based on user feedback. To provide the best and smooth user experience, it’s important that the AI technology blends seamlessly with the product.